Protein Secretion in Plants: from the trans-Golgi Network to the Outer Space


*Corresponding author: Gerd Jürgens,


Functional analysis of exocytosis in yeast and animal cells has led to the identification of conserved elements and mechanisms of the trafficking machinery over the last decade. Although functional studies of protein secretion in plants are still fairly limited, the Arabidopsis genome sequence provides an opportunity to identify key players of vesicle trafficking that are conserved across the eukaryotic kingdoms. Here, we review and add to recent genome analyses of trafficking components and highlight some plant-specific modifications of the common eukaryotic machinery. Furthermore, we discuss the evidence for targeted, polarised secretion in plant cells, and speculate about possible underlying cargo sorting processes at the trans-Golgi network and endosomes, based on what is known in animals and yeast.

Protein delivery to the cell surface, via the endomembrane system, is a common feature of eukaryotic cells. Integral proteins of the plasma membrane as well as secreted proteins are synthesised at the endoplasmic reticulum (ER) and are inserted into or translocated across the ER membrane. All subsequent steps of protein trafficking involve transport vesicles that bud from a donor membrane and fuse with a target membrane. Most vesicles that leave the ER are trafficked to the Golgi complex, although some vesicles bypass the Golgi on their way to the vacuole (1). The Golgi complex is a major sorting station where trafficking routes diverge to the plasma membrane, the endosome, the vacuole or the cell plate (2) (Figure 1). Post-Golgi sorting occurs in the endosome which is also involved in recycling of vesicles to the plasma membrane.

Figure 1.

Model of the trafficking routes in the late secretory system. Schematic of an Arabidopsis root meristem cell expressing plasma membrane markers located to distinct subdomains (marker names are coloured and in italic). Coloured arrows mark the respective possible pathways to different plasma membrane compartments from the TGN and/or the endosome. The size of the arrows indicates the possible differences in relative contributions to the total transport of these plasma membrane markers. Names of membrane compartments are in bold. Reported locations of proteins of the transport machinery mentioned are in plain text. The circle of dotted lines represents the ill-defined plant endosome, whose structure and subcompartments (early endosomes, sorting endosomes, etc.) will have to be defined in the future.

Extensive analysis of protein secretion in yeast and animal cells suggests that the basic machinery was invented by a unicellular eukaryote and subsequently adapted to the needs of multicellular life. For example, targeted secretion to the apical vs. baso-lateral surface of polarised epithelial cells requires sorting of proteins destined to specific subdomains within the plasma membrane and uses modified subsets of trafficking components already present in yeast. Compared to yeast and animals, much less is known about mechanisms of protein secretion in plants (2). Considering that plants and animals display different cellular organisations and also have achieved multicellularity independently during evolution, it seems likely that protein secretion has undergone specific modifications in the plant lineage.

A few examples from plant development may illustrate requirements that have to be met by the protein secretory system. Vegetative pollen cells secrete a peptide ligand (SCR) into their cell wall that interacts with a Ser/Thr kinase receptor complex (SRK-SLG) in the plasma membrane of stigmatic cells, ensuring recognition of matching partners in the self-incompatibility response (3). Similarly, signaling between the stem cells and the organising centre of the shoot meristem requires interaction between a secreted peptide ligand (CLV3) and a Ser/Thr kinase receptor complex (CLV1-CLV2) for continual reassessment of cell fate and maintenance of shoot meristem organisation (4). Whereas secretion of the ligand into the apoplast has been demonstrated, internalisation of the ligand-receptor complex and potential recycling of the receptor complex to the plasma membrane have not been studied. Root hairs and pollen tubes grow by delivery of transport vesicles to their tips (5). Moreover, pollen tubes grow directionally towards the ovules for fertilisation, which requires continual changes of vesicle targeting to the plasma membrane in response to as yet unknown chemotactic signals (6). Plant cells are able to elongate directionally up to 200 times their original length (7), which requires differential insertion of membrane material into subdomains of the plasma membrane. Some cell types such as trichomes and root hairs undergo localised outgrowth from specific sites within the plasma membrane (8,9). Finally, the cells of the endodermis partition their surface into two domains, with an inner surface separated from an outer surface by the Casparian strip (10). This is formally similar to the zonula adherens border between apical and baso-lateral domains of animal epithelial cells.

The cellular organisation of plants may entail specific modifications of protein secretion. Not only does an ER network occur at the cell periphery, but also the Golgi complex is dispersed into many stacks whose numbers vary from about 20 to 400 in plant cells (11). Furthermore, the endomembrane organisation is highly dynamic, due to extensive cytoplasmic streaming, which moves organelles about the cell and may bring Golgi stacks and ER in close proximity (12,13). As plant cells enlarge, they often contain a large central vacuole that leaves a thin layer of cytoplasm underneath the plasma membrane. Specific cell types contain two different kinds of vacuole: a lytic vacuole akin to those in yeast or the lysosome of animal cells and a storage vacuole from which reserve materials can be mobilised (14). The plant cytoskeleton also has unique features. Actin filaments mediate cell growth, cytoplasmic streaming, ER–Golgi association and organelle movement. Microtubules are not nucleated from localised MTOCs (centrosomes) and form particular arrays such as interphase cortical hoops that maintain cell shape, and preprophase band and phragmoplast that mediate oriented cell division and the execution of cytokinesis, respectively (15). However, MTs do not seem to be necessary for basic cell growth, as protein trafficking to the plasma membrane occurs in MT-deficient cells (16). Moreover, the extracellular matrix (ECM) of plant cells is a largely polysaccharidic cell wall, whereas the animal ECM is mainly proteinaceous. Finally, due to the absence of cell movement, plant organs are shaped by oriented cell division and regulated cell expansion, with the latter resulting from targeted protein secretion.

This review serves two purposes. First, we will exploit the Arabidopsis genome sequence to provide an overview of some central players of the plant secretion machinery by comparison with the components that have been functionally characterised in animals and yeast. Second, we will summarise recent findings that may shed light on protein trafficking between the trans-Golgi network (TGN) and the plasma membrane, discussing the role of the endosomal system in protein sorting. Trafficking from the TGN to the vacuole has been reviewed recently (17) and will not be covered here.

Plant Proteins Involved in Vesicle Trafficking

Studies in yeast and animals have characterised components of an ever-growing molecular machinery that ensures the budding of vesicles with defined sets of cargo proteins from the donor membrane as well as their recognition by, and fusion with, the correct target membrane. Screening of the Arabidopsis genome for related sequences to key players of intracellular trafficking in nonplant organisms gives some idea of conserved elements of the machinery between yeast, plants and animals. For the sake of brevity, we will limit our discussion to two processes: vesicle budding from the donor membrane and vesicle interaction with the target membrane.

Vesicle Budding

Vesicle budding requires small GTPases of the ARF family, their guanine-nucleotide exchange factors (ARF-GEFs) and GTPase-activating proteins (ARF-GAPs) for coat recruitment and cargo selection. ARFs not only act to recruit COPI and clathrin coats to membranes but also play a role in the control of membrane lipid composition, actin remodeling and related events (18). Structurally related ARF-like proteins (ARLs) are members of the same subgroup of the ras superfamily, but have different functions. For example, Arabidopsis ARL2 is involved in microtubule formation (16). The six mammalian ARFs have been grouped into three classes based on their structure and function. Class I consists of three members (ARF1–3), class II is represented by ARF4 and ARF5, whereas ARF6 constitutes the most divergent class III (19). By contrast, there are only three ARFs in yeast. Class I ARF1 and ARF2 are functionally redundant (20). Yeast ARF3 is divergent from both yeast class I ARFs and human ARF6 (21). In Arabidopsis, six of nine putative ARF genes encode class I ARF proteins with 98–100% amino acid identity (Table 1). Four of them can be grouped into two pairs derived from segmental genome duplication events. Additionally, there are three more divergent ARFs, including one pair of duplicates, with about 60% amino acid identity to human ARF1. The third divergent ARF seems to represent a subclass of its own (Table 1). The less conserved Arabidopsis ARFs are also divergent from human ARF6 or yeast ARF3, arguing for an independent evolution of divergent ARF classes from class I ancestors in the three kingdoms.

Table 1. : Overview of Arabidopsis thaliana protein families involved in vesicle trafficking. Asterisks indicate searches done by the authors
Protein groupSubgroup/classProteins/
protein numbers
AGI numbersRef.
ARFclass IARF1 – ARF6At3g62290; At2g47170; At1g23490;*
G-proteins  At1g70490; At1g10630; At5g14670 
plant class AARF7 – ARF8At5g17060; At3g03120*
plant class BARF9At2g15310*
ARF-GEFsGea/GNOM/GBFGNOM, GNL1, GNL2At1g13980; At5g39500; At5g19610*
Sec7p/BIG classAtBIG1-5At3g43300; At1g01960; At3g60860;
At4g35380; At4g38200
ARF-GAPsARF-GAP1 class3 membersAt2g35210; At4g17890; At5g46750*
Age2p class2 membersAt3g17660; At5g54310*
Age2p-like class3 membersAt3g07940; At4g05330; At4g21160*
Gcs1p class2 membersAt2g37550; At3g53710*
Rab GTPases12 subfamilies57 membersSee reference(44)
Rab effectorsexocyst8 of 8 subunitsAt1g47550, At3g10380, At1g71820
(At1g21170/At1g76850), At5g03540
(At1g10385/At5g49830), At5g12370
VFT3 of 3 subunits(At1g71270/At1g71300), At1g50500,
HOPS6 of 6 subunitsAt2g05170, At2g38020 (VCL1),
At1g12470, At1g08190, At4g36630,
Sec34/35p5 of 8 subunitsSee reference(52)
SNAREsSYP (syntaxins)24 membersSee reference(57)
VAMP14 members  
SNAP25 hom.3 members  
NPSN3 members  
 VTI, Gos1,
Bet1, Membrin
3, 2, 2, 2 members,
Sec1 familySec1 groupKEULE, Sec1a, Sec1bSee reference(57)
Vps45 groupVPS45  
Vps33 groupVPS33  
Sly1 groupSly1  

The ARF GDP/GTP exchange factors (ARF-GEFs) also underwent diversification during the separate evolution to multicellularity. Yeast have four such regulators of ARFs: the largely redundant couple Gea1p/Gea2p, and Sec7p and Syt1p (18). ARF-GEFs are defined by the catalytic Sec7 domain. The mammalian cell displays a much greater diversity of Sec7 domain proteins than yeast. There are to date five different classes of exchange factors described, accounting for about 10 different ARF-GEFs three of which show a completely new arrangement of domains that are not present in yeast ARF-GEFs (22,23). These new ARF-GEF classes appear to have evolved for specialised transport functions in a multicellular context, such as matrix adhesion or receptor down-regulation (24,25).

Plant ARF-GEFs appear to show a different pattern of diversification. Arabidopsis has only two classes of exchange factors which can be distinguished by the presence or absence of an N-terminal dimerisation domain and are related to yeast Gea1/2p and Sec7p large ARF-GEFs, respectively (22,26). Unlike mammals with only one Gea1/2p class (GBF) and two Sec7p class (BIG1/2) ARF-GEFs, the Arabidopsis genome encodes 3 Gea1/2p and 5 Sec7p-like exchange factors. The sequence divergence between family members (20–60% identity) suggests at least partially non-redundant functions. Only the Gea1/2p class ARF-GEF GNOM has been functionally characterised. Although gnom embryos have severe patterning defects, the mutant cells are viable and can be grown in culture (27–29). GNOM might act in signal-dependent recycling of plasma membrane proteins from endosomal compartments to the cell surface (29) (Figure 1 and see below). Neither yeast nor mammalian ARF-GEFs of the Gea/GBF/GNOM class appear to act in endosomal recycling, which in mammals rather involves mammal-specific ARF-GEFs, such as EFA6, ARNO or ARF-GEP100 (22,23). Thus, plants seem to have evolved new members of an existing ARF-GEF class to regulate recycling events, whereas ARF-GEFs with newly assembled domain structures have evolved for the same process in animals.

Yeast ARF1-GAP mediates interaction between ER-Golgi v-SNAREs and COPI coat proteins (30). Three other yeast ARF-GAPs act at the TGN, whereas mammals also have structurally divergent ARF-GAPs functioning in the periphery of the cell (18,31). In Arabidopsis, ten genes encode ARF-GAPs of which three groups are related to the yeast ARF-GAPs acting at the TGN, Age2p, Gcs1p and Glo3p, respectively, while the members of the fourth group show similarities to human ARF-GAP1 (31,32) (Table 1).

Transport vesicles can be distinguished by their coats which are formed by a limited set of proteins, including COPI, COPII and clathrin. Coat proteins are captured by specific adaptor proteins and aid in cargo recruitment and vesicle budding (33). Whereas COPI- and COPII-coated vesicles traffic between ER and cis-Golgi or between Golgi cisternae, clathrin-coated vesicles bud from the TGN or the plasma membrane and are thus involved in post-Golgi trafficking.

Clathrin-coated vesicles have been detected at the TGN and plasma membrane, and at the cell plate in dividing cells of Arabidopsis(34–36). Arabidopsis has two genes encoding slightly divergent putative clathrin heavy chains, and clathrin has been immunolocalised to budding vesicles (35). In addition, one of three putative clathrin light chains was shown to interact with light chain-free mammalian clathrin hubs in vitro (37).

Clathrin is recruited by adaptor (AP) complexes that consist of four subunits: two large subunits (β and one of α, γ, δ or ε), the medium subunit μ and the small subunit σ. Like mammals, and in contrast to Drosophila or yeast, Arabidopsis has four adaptor complexes, AP-1 to AP-4 (38). AP-2 mediates clathrin recruitment in endocytosis from the plasma membrane, whereas AP-1, AP-3 and AP-4 are associated with the trans-Golgi network and/or endosomes in mammals. Whereas AP-1 and AP-2 function with clathrin, AP-4 is most likely part of a nonclathrin coat. Interestingly, the Arabidopsis genome encodes three putative γ subunits of AP-1 as opposed to two in mammals and only in one each in Drosophila and yeast, and three β subunits that cannot be clearly assigned to either AP-1 or AP-2 (38,39). The reason for this potential diversity of AP-1 and, possibly, AP-2 complexes remains to be determined as no functional data are available in Arabidopsis.

Clathrin adaptors recruit cargo proteins via interaction with their cytosolic tails, e.g. μ2-adaptin of AP-2 recognises a tyrosine-based endocytic sorting motif (40). In plants, the only evidence for cargo recruitment is the in vitro interaction of the TGN-localised vacuolar cargo receptor AtELP with the mammalian TGN-specific AP-1 clathrin-adaptor complex (41,42).

Interaction of Vesicles with the Target Membrane

The fusion of transport vesicles with their target membrane is preceded by tethering and docking, both of which contribute to the target specificity of vesicle fusion. Initially, rab-GTP on the vesicle membrane interacts with effectors that tether the vesicle to the target membrane (43). This is followed by the formation of trans-SNARE complexes that dock vesicles to target membranes, facilitating subsequent fusion.

Rab proteins are thought to determine the fusion competence of membranes and to specify target membranes by acting as molecular scaffolds and regulators of membrane composition. As many as 57 rabs have been identified in Arabidopsis, nearly matching the number of rabs in mammals and by far exceeding those of yeast, Drosophila or Caenorhabditis elegans (44). Rab functions are conserved across eukaryotes, such that their subcellular localisation can be inferred from known localisations of members of the same subfamily in other species. The large number of Arabidopsis rab proteins results from diversification within 12 subfamilies. The rab11 subfamily, for example, contains 26 members, which may reflect differential tissue expression or functional redundancy of late-endosomal rabs. Alternatively, this diversification might suggest a complex subdomain structure of recycling endosomes in plants [for review, see (43)]. The rab11 homologue PRA2 has been reported to localise to the ER and to regulate brassinosteroid biosynthesis in response to light, an unprecedented case of a rab function possibly unrelated to membrane trafficking (45). However, the same protein was recently localised to the Golgi and endosomes rather than the ER (46) (Figure 1). Regardless of this discrepancy, both reports agree on differences in localisation or function between PRA2 and its closest homologue PRA3, supporting the notion of functional diversity among rab11 subfamily members in plants.

Another example of a plant-specific rab variant is Ara6, a member of the rab5 subfamily, which lacks the strictly conserved C-terminal isoprenylation (47). Instead, N-myristoylation and palmitoylation of an N-terminal extension is necessary for proper localisation of Ara6 to endosomal compartments (Figure 1). Since only a subset of early endosomes appears to be Ara6-positive, plants may have functionally distinct early endosome populations.

Rab effectors are specific to the target membrane and often consist of protein complexes. For example, the exocyst complex of eight proteins conserved between yeast and mammals mediates tethering of Sec4p/rab3 vesicles to the plasma membrane (48). In yeast, the exocyst targets vesicles to a specific site of the bud plasma membrane via the Sec3p protein, although t-SNAREs appear more broadly distributed in the target membrane (49). The mammalian exocyst targets vesicles to the baso-lateral surface of epithelial cells (50). A Ypt6p effector complex named VFT (for Vps 53) mediates tethering of putatively endosome-derived vesicles to the trans-Golgi network (51). A Ypt1p effector complex named Sec34/35 complex appears to tether vesicles to the cis-Golgi stack (52). Interestingly, several components of these three complexes show similarities that suggest a common evolutionary origin (53). The Arabidopsis genome encodes homologues of all eight exocyst, three VFT and five of eight Sec34/35 complex proteins (Table 1). However, no functional analyses have been performed. Vesicles trafficking to the vacuole are tethered to the target membrane via the Ypt7p effector complex named HOPS, which is also conserved between yeast and mammals (54). Again, homologues of all six components of the HOPS complex are encoded in the Arabidopsis genome (Table 1). In this case, mutations in the VACUOLELESS1 (VCL1) gene encoding the Vps16p homologue were shown to be embryonic lethal and to lead to defects in vacuole biogenesis (55).

Vesicle docking is brought about by the pairing of complementary SNAREs on opposite membranes (56). SNARE complexes consist of pathway-specific synaptobrevins (v-SNAREs) and syntaxins (t-SNAREs), whereas more promiscuous t-SNARE light chains or SNAP25s contribute the remaining two coiled-coil domains of the four-helical bundle in yeast and mammals. Again, the Arabidopsis genome encodes a larger number of syntaxins (called SYP, syntaxins of plant) than does the human genome (57). Especially, there are nine SYP1 proteins related to the plasma-membrane syntaxins Sso1/2p and syntaxin1. They include KNOLLE, a cytokinesis-specific syntaxin only found in plants, and the plasma membrane-located Syr-1 required for protein secretion (58–60) (Figure 1). It will be interesting to determine the tissue specificity and the subcellular location of the remaining members of this group. Members of four other SYP subfamilies form at least five complexes involved in TGN/prevacuolar compartment (PVC) trafficking (61). Members of the SYP2 and SYP4 syntaxin family are closely related by sequence, but SYP41 and SYP42 have been shown to localise to different regions of the TGN. In addition, functional analysis by reverse genetics suggests unique requirements of each, as evidenced by pollen lethality (62). As in yeast and animals, the SNAP25-homologue SNAP33 appears to be more promiscuous, interacting with KNOLLE at the cell plate and also localising to the plasma membrane (63) (Figure 1).

Arabidopsis encodes 16 putative synaptobrevins, of which 11 form a subfamily related to the endosomal/lysosomal VAMP7 (57). By contrast, there are no close homologuesto the plasma membrane synaptobrevins Snc1/2p and VAMP1/2. On the other hand, the Arabidopsis genome encodes 3 ‘novel plant SNAREs’ (NPSNs) that lack close homologues in nonplant organisms. NPSN11 interacts with the cytokinesis-specific syntaxin KNOLLE and also localises to the cell plate during cytokinesis (64). It remains to be determined whether NPSNs can substitute for conventional synaptobrevins.

Arabidopsis encodes six putative members of the Sec1 family which in nonplant organisms interact with SNAREs in different ways. Whereas nSec1 stabilises the closed conformation of syntaxin1A (65), Sly1 interaction with the N-terminus of the cis-Golgi SNARE Sed5 may contribute to the specificity of SNARE complex formation (66,67). In contrast to yeast, the Arabidopsis VPS45 is located at the TGN and interacts with the TGN syntaxins of the SYP4 family (68) (Figure 1). The only Arabidopsis Sec1 protein that has been functionally characterised is KEULE which interacts with the cytokinesis-specific syntaxin KNOLLE and is required for vesicle fusion during cell-plate formation (69,70).

In conclusion, the sequence analysis of the Arabidopsis genome suggests conservation of key regulators of vesicle trafficking. However, in the case of protein families, there has been differential expansion of subfamilies in plant evolution, which points to possibly divergent roles of individual members. To clarify this issue, functional analysis needs to be done by reverse genetics in Arabidopsis.

Post-Golgi Trafficking Pathways in Plants: Where Does Sorting Take Place?

Protein secretion to the cell surface delivers integral plasma membrane proteins, such as cellulose synthases, the t-SNAREs SYR1 and SNAP33 or H+-ATPase, and extracellular proteins located in the cell wall, such as lipid transfer protein, endoxyloglucan transferase or endoglucanase, and secreted signaling peptides, such as CLV3 or SCR. In all these cases, the entire plasma membrane is the target of vesicle trafficking. However, there are other proteins that accumulate in specific subdomains of the plasma membrane. This is illustrated by the apical localisation of the auxin uptake carrier AUX1 (71), the basal localisation of the putative auxin efflux carrier PIN1 (28,72) and the lateral localisation of COBRA, a GPI-anchored protein necessary for correct differential cell elongation (73) (Figure 1). These three proteins can be found in the same cells of the Arabidopsis seedling root, suggesting that targeted secretion can distinguish between different subdomains of the plasma membrane. Other proteins localised to specific subdomains include PIN2, which accumulates at the apical end of root epidermal cells (74), and PIN3, which is predominantly observed at lateral plasma membranes of root cortical cells (75).

How these differential localisations are achieved by targeted secretion in plant cells is not known. In mammalian polarised epithelial cells, syntaxins are differentially localised to either apical or basolateral plasma membrane domains (76), and this was shown to be important for correct sorting of marker proteins (77). It is important to note that epithelial cells have mechanical diffusion barriers that prevent correctly targeted proteins from diffusing laterally into the neighbouring plasma-membrane domain. Such barriers are not apparent in the depicted root meristem cell (Figure 1), suggesting that different mechanisms prevent the mixing of apical, lateral and basal markers. One possible mechanism would be a self-organising scaffolding machinery. Alternatively or in addition, segregated proteins might be continually resorted by recycling, as will be discussed below. However, where does sorting happen in the first place?

The TGN has been recognised as a major sorting site in the secretory pathway where proteins destined for the lysosome/vacuole are segregated away from exocytic cargo. This is also the case in plants since the vacuolar sorting receptor AtELP has been localised to the TGN (41,42) (Figure 1). It has also been shown recently through live imaging that transport vesicles from the TGN directly fuse with the plasma membrane, without passing through any intermediate compartment (78). If the TGN acts as a last sorting station before outer space, differential secretion towards plasma membrane subdomains would require budding of several distinct vesicle populations from the TGN. However, there is evidence from both mammalian and yeast cells for ‘indirect’ trafficking to the plasma membrane, with vesicles first being delivered to the endosomal system (79–82). The endosome is an important sorting centre in the cell where cargo destined for degradation is separated from proteins to be recycled back to the surface. Thus, it is a rather straightforward idea that newly synthesised proteins travel from the TGN to the endosome where they join proteins from the endocytic pathway, and both sets of proteins are sorted together for their respective destinations at this point. The direct pathway from the TGN to the plasma membrane could be taken by cargo such as cell-wall material or secreted proteins that would not undergo recycling through endosomes. In a growing plant cell, this would probably still constitute the major part of vesicle flow to the plasma membrane. However, things are likely to be more complicated, since sorting of some apical or basolateral markers does clearly also take place in the TGN. This is even the case in nonpolarised cells, where different vesicle populations are not destined to distinct plasma membrane compartments (78). Currently, data from mammals and yeast only indicate the existence of two pathways from the TGN to the plasma membrane, but we are far from understanding why one protein takes the direct route and another travels via the endosome, and what contributions the two pathways make to overall secretion.

In plants, post-Golgi trafficking to the plasma membrane has not been analysed and it is thus not clear whether there are multiple pathways. However, BFA treatment leads to rapid internalisation of plasma membrane proteins, which is most likely due to a block of re-secretion to the plasma membrane (29). BFA-induced internal accumulation appears to be plant specific, since in animals endocytosis and recycling of plasma membrane markers are resistant to BFA (83,84). Thus, BFA is a tool to study post-Golgi secretion in plants. Several plasma membrane proteins accumulate intracellularly in response to BFA treatment (28,29,63,85). It has been shown that PIN1 accumulates in so-called BFA compartments that consist of aggregates of membrane vesicles and are distinct from adjacent Golgi stacks (29). The BFA compartments involved in PIN1 recycling most likely represent endosomes. If this is the case, proper targeting of PIN1 and possibly other plasma membrane proteins may be effected by sorting in endosomes rather than at the TGN. By comparison, newly synthesised PIN1 accounts only for a minor proportion of transported PIN1 protein in the cell. Thus, regardless of whether this minor fraction is sorted at the TGN or passed on to the endosome for sorting, the determining step for the polar localisation of PIN1 should be sorting at the endosome. How sorting and targeting to the correct plasma membrane subdomain is brought about is entirely unknown. For example, none of the several putative plasma membrane SNAREs of Arabidopsis has been shown to be asymmetrically localised in the plasma membrane, in contrast to animals (76). In the case of the COBRA protein, which is localised to the lateral subdomain of the plasma membrane, its GPI anchor may be instructive for targeting since, in animals, segregation into lipid rafts has been postulated to target GPI-anchored proteins to the apical end of epithelial cells (86).

Future Perspectives

Both the analysis of the Arabidopsis genome and the still limited number of functional studies point to a similar complexity of the secretory system in plants as in animals. More functional studies are required to analyse mechanisms of plant protein secretion, and the necessary tools are now becoming available. Specifically, reverse-genetics analysis of Arabidopsis genes has become commonplace with the availability of large collections of insertion lines, and facile transformation makes it easy to express specific variants of regulatory components of vesicle trafficking. In addition, the paucity of suitable subcellular markers for identifying membrane compartments will be overcome by the widespread use of GFP fusions. We can thus expect a dramatic increase in knowledge about protein secretion in plants, which will eventually enable us to differentiate between a common eukaryotic heritage and plant-specific ways of sorting out things.